{"paper":{"title":"LNN-PINN: A Unified Physics-Only Training Framework with Liquid Residual Blocks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Adding liquid residual gating inside hidden layers improves PINN accuracy on benchmarks while keeping training unchanged.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Fujun Liu, Hanxuan Wang, Ze Tao","submitted_at":"2025-08-12T13:35:46Z","abstract_excerpt":"Physics-informed neural networks (PINNs) have attracted considerable attention for their ability to integrate partial differential equation priors into deep learning frameworks; however, they often exhibit limited predictive accuracy when applied to complex problems. To address this issue, we propose LNN-PINN, a physics-informed neural network framework that incorporates a liquid residual gating architecture while preserving the original physics modeling and optimization pipeline to improve predictive accuracy. The method introduces a lightweight gating mechanism solely within the hidden-layer"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across four benchmark problems, LNN-PINN consistently reduced RMSE and MAE under identical training conditions, with absolute error plots further confirming its accuracy gains.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The observed accuracy improvements arise solely from the architectural addition of the liquid residual gating mechanism inside the hidden-layer mapping and not from any unintended change in effective capacity, optimization dynamics, or data handling.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"LNN-PINN integrates liquid residual blocks into PINNs and reports lower RMSE and MAE on four benchmark problems while leaving the original physics modeling and optimization pipeline unchanged.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Adding liquid residual gating inside hidden layers improves PINN accuracy on benchmarks while keeping training unchanged.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e25485531dc03187fb34dd0c6453f29cfdf17b6b1e236286f0dab0f0af2a37fb"},"source":{"id":"2508.08935","kind":"arxiv","version":4},"verdict":{"id":"57ec6584-2692-48cb-be6b-61cb2cdf25c8","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-18T23:31:27.373019Z","strongest_claim":"Across four benchmark problems, LNN-PINN consistently reduced RMSE and MAE under identical training conditions, with absolute error plots further confirming its accuracy gains.","one_line_summary":"LNN-PINN integrates liquid residual blocks into PINNs and reports lower RMSE and MAE on four benchmark problems while leaving the original physics modeling and optimization pipeline unchanged.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The observed accuracy improvements arise solely from the architectural addition of the liquid residual gating mechanism inside the hidden-layer mapping and not from any unintended change in effective capacity, optimization dynamics, or data handling.","pith_extraction_headline":"Adding liquid residual gating inside hidden layers improves PINN accuracy on benchmarks while keeping training unchanged."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.08935/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":30,"sample":[{"doi":"","year":2019,"title":"Physics- informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations","work_id":"c062108e-56fd-4e26-b54a-22a7eb4bec24","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Physics-informed neuralnetworksforpdeproblems:acomprehensivereview","work_id":"13b3acf2-9351-45e8-8a24-f8b67a6d28a3","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Physics-informed neural networks(pinn)forcomputationalsolidmechanics:Numericalframe- works and applications.Thin-Walled Structures, 205:112495, 2024","work_id":"dea7c9e4-f08e-4be5-9f3c-b88370910c29","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Analytical and neural network approaches for solving two-dimensional nonlinear transient heat conduction","work_id":"cd0e3f9b-a976-4d7c-8239-76db218e49ae","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Physics- informed neural networks with adaptive localized artificial viscosity","work_id":"59c73513-918e-477c-b141-120de7aeace4","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":30,"snapshot_sha256":"cfdc290041da2eb2d4fa86551063ec2e4613f0f161eeb501fe6775d83da6a814","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}